Prediction Of Aeration Capacity Of Biochemical Tank For Wastewater Treatment Based On Machine Learning

In modern times,with the continuous updating and integration of technology,intelligent water has become the development trend of wastewater treatment industry.New technologies such as Internet of Things,big data,and intelligent algorithm model continue to enter the traditional industry and constantly intergrate with the traditional.For urban wastewater treatment plant,biochemical reaction tank is the core of the whole process and also the largest energy consumption tank of the whole process,in which the electric energy consumed by aeration accounts for 50%-75%of the total electricity energy.On the one hand,the quality of aeration determines the efficiency of biochemical reaction,and affects the degradation of organic pollutants and the effect of nitrogen and phosphorus removal.On the other hand,it also significantly influences the standard of effluent water quality.There is a need to save energy and reduce redundancy while ensuring that the water quality of the effluent is up to standard.How to accurately aerate is a difficult problem at present.In this paper,an aeration capacity prediction method based on machine learning is proposed by combining the intelligent models with the biochemical tank aeration capacity prediction.The main results are as follows.1)Starting from the actual wastewater treatment plant,the data of actual water quality and quantity of inlet and outlet of biochemical tank collected on site are analyzed in detail.The content,scope and general dynamic change of the index are studied,and the current deficiencies are discovered,that is,the effluent ammonia nitrogen concentration is far below the standard.Although it meets the criteria,this is the result of consuming more aeration,which is the outcome of redundant aeration.2)After exploring the availability of the data collected on site,the real data are selected to build the random forest,extremely randomized tree,gradient boosted decision tree,extreme gradient boosting tree,light gradient boosting machine and ridge regression models to fit the relationship between the water quality inlet and outlet of the biochemical tank with the amount of aeration respectively.The best fitting effect is obtained by extremely randomized tree model(R~2=0.8418),and the weakest model is ridge regression model(R~2=0.3891).Although most of the models have certain learning ability,there are still some deficiencies.At the same time,the reasonableness of aeration capacity prediction is explained by adjusting ammonia nitrogen concentration of effluent.3)The model fusion and optimization of the Stacking method are selected.The six models above are integrated to further predict the biochemical tank aeration,and the fitting effect is improved(R~2=0.8679).The results show that the optimized model can better predict the required aeration capacity of the biochemical tank based on the data of real-time incoming and outgoing water quality and quantity.At the same time,the important factors affecting the aeration rate are analyzed.According to the results of importance ranking,its significances in the actual process flow are put forward.The optimization results of the prediction method in the actual process are explained as well.In summary,based on the limited real-time water quality and quantity data of thebiochemical tank at this wastewater plant,an intelligent simulation model has been constructed to enable accurate prediction of aeration volumes resulting in effectively saving energy and safeguarding the carbon neutral and sustainable development of the urban wastewater treatment plant.